1

SELECTING AUSTRALIAN EQUITY SUPERANNUATION FUNDS: A RETAIL INVESTOR’S PERSPECTIVE

Michael E. Drewa, Jon D. Stanfordb* and Madhu Veeraraghavanc

a School of Economics and Finance Queensland University of Technology

GPO Box 2434 Brisbane Queensland 4001

Australia

b School of Economics The University of Queensland

Brisbane Queensland 4072 Australia

c School of Accounting and Finance

Griffith University Gold Coast Campus

PMB 50 GCMC Queensland 9726

Australia

November 2002

Discussion Paper No 320

ISSN 1446-5523

Drew, Stanford & Veeraraghavan. This discussion paper should not be quoted or reproduced in whole or in part without the written consent of the author.

2

1. INTRODUCTION

DOES ACTIVE INVESTMENT MANAGEMENT ADD VALUE? This question has been the source of

continued debate in financial economics since the contributions of Sharpe (1966), Treynor

(1966) and Jensen (1968). One strand of literature finds that investment managers have little

stock-selection ability consistent with the efficient market framework of Fama (1970). Research

from the United States by Malkiel (1995) and Gruber (1996) and the United Kingdom by Leger

(1997) finds that the investment management industry, on average, destroys value for investors

through under-performing benchmark returns. For instance, Gruber (1996) reports that the

average mutual fund under-performs index returns by some 65 basis points per annum for the

period 1985 through 1994. These studies advocate a passive approach to the stock-selection

problem.

By contrast, another strand of literature finds some limited evidence of stock-selection ability by

managers. Grinblatt and Titman (1989), Wermers (2000) and Kosowski, Timmermann,

Wermers and White (2001) find that investment managers select stocks that outperform

benchmark returns, reflecting the incomplete arbitrage model of Grossman and Stiglitz (1980).

Moskowitz (2000) explains that this second set of studies examines the individual equity

holdings of funds, creating a hypothetical portfolio for each fund that contains only stocks and

does not account for transaction costs or expenses. Wermers (2000) reports that while the gross

returns from equity holdings outperform a broad market index by 130 basis points per year, the

net fund returns under-perform the same index by 100 basis points per year1.

1 Wermers (2000) reports that of this 230 basis points difference, approximately 160 basis points is split evenly between fund expenses and transaction costs, with the remainder attributed to bond and cash holdings.

3

While both schools take differing approaches to the evaluation problem, considerable consensus

is found that, as an industry, investment managers under-perform stated benchmarks on an after-

fee basis (i.e., post transaction costs and management expenses). The value (or otherwise) of

active management has immediate implications for Australia’s system of retirement funding,

termed superannuation2. This is potentially important as the international experience suggests

that active investing resulted in a high-cost production function for investment management that

yields, in aggregate, poor results. Investors that select an active fund require the manager to

execute stock trades at prices sufficiently different from fully-informed prices to, first,

compensate them for the cost of becoming informed and, second, to earn superior risk-adjusted

returns.

Recent research considering this issue by Drew and Noland (2000), Drew and Stanford (2000,

2001a, 2003), Sawicki (2000) and Sawicki and Ong (2000) for the Australian setting provides

corroborating evidence of the experience in the United States. Using techniques comparable to

Gruber (1996), Drew and Stanford (2000) find that the average domestic equity superannuation

fund under-performs benchmark returns by a range of 46 to 93 basis points per annum for the

period 1991 through 19993. Drew and Stanford (2000) find that as an industry, investment

managers destroy value for superannuation members, with the costs of research and trading

associated with active management being largely sunk.

2 A common type of managed fund in Australia is the superannuation fund. Superannuation funds are designed to set aside an amount during the working lives of people so as to meet their financial needs during retirement. 3 Moreover, Drew and Stanford (2001b) find that active funds are regularly terminated due to poor performance, with survivorship bias negatively affecting industry performance by 23 basis points per annum on a risk-adjusted basis.

4

This study departs with the tradition of a broad industry based evaluation of the investment

management industry and considers whether the portfolio returns achieved by individual

investment managers persist through time. Are there investment managers that have a ‘hot hand’

providing consistently high returns for investors? Can investors fashion a fund selection strategy

that would, ex-ante, permit them to garner superior returns? Specifically, this study tests the

hypothesis that the relative return achieved by a fund last year has no predictive value for

tomorrow.

2. PREVIOUS RESEARCH

The ability to predict the future performance of funds based on ex-ante information has been the

topic of intense debate by investors, practitioners and researchers alike. The received statement

of market efficiency, the efficient market hypothesis, implies that historical performance is no

guide to future performance and that any excess performance achieved by an investment

manager is the result of chance, not the skillful application of active stock selection techniques.

However, empirical testing of this position has provided mixed results over the 1990s.

The persistence case, forwarded by Hendricks, Patel and Zeckhauser (1993), Goetzmann and

Ibbotson (1994), Brown and Goetzmann (1995), Kahn and Rudd (1995), Bal and Leger (1996),

Elton, Gruber and Blake (1996b), Gruber (1996)4, Stewart (1998), and Carpenter and Lynch

(1999) report that past returns and relative rankings are useful in predicting returns and rankings

4 Gruber (1996) finds that expenses, raw returns, risk-adjusted alphas, multifactor asset pricing model alphas and new money flows into mutual funds forecast positive relative performance.

5

in the short run (1 to 3 years). Grinblatt and Titman (1992) provide longer-term evidence of

persistence (10 years), indicating that there is positive persistence in mutual fund performance.

In a novel approach, Bauman and Miller (1994, 1995) rank the performance of funds over

complete stock market cycles, reporting that the correlations of portfolio performance rankings

from one market cycle to the next are generally positive and meaningful. Gruber (1996)

concludes that these results are of “economic importance (pp.795)” with investors buying the top

ranked funds from previous periods earning superior raw and risk-adjusted returns in the future.

If the persistence anomaly holds, superannuation investors could achieve their retirement

objectives far more rapidly through the selection of active managers, on an ex-ante basis, that

would consistently deliver superior returns.

The case rejecting the differential skill of managers consistently through time is led by Troutman

(1991), Brown, Goetzmann, Ibbotson and Ross (1992), Lakonishok, Shleifer and Vishny (1992),

Bogle (1995), Malkiel (1995), Carhart (1997), Cheng, Pi and Wort (1999). These researchers

consider both raw and risk-adjusted returns from individual funds and investment management

firms, answering the question of whether persistence is economically significant in the negative.

The contribution of Carhart (1997) shows the failure of the capital asset pricing model to capture

the cross-section of fund returns (particularly relating to short-term momentum effects in stock

returns) is responsible for the persistence puzzle. Carhart (1997) observes that performance

persistence is simply a matter of luck, stating that “(these funds) accidentally end up holding last

year’s winners. Since the returns on these stocks are above average in the ensuing year, if these

funds simply hold their winning stocks, they will enjoy higher one-year expected returns and

6

incur no additional transaction costs for this portfolio. With so many mutual funds, it does not

seem unlikely that some funds will be holding many of last year’s winning stocks simply by

chance (pp. 73).” This study adjusts for risk using the Sharpe (1966, 1994) index, in an attempt

to mitigate the problem of benchmark inefficiency.

Arteaga, Ciccotello and Grant (1998) find that performance persistence by investment

management firms is captured by marketing oriented explanations. For instance, Arteaga et al

(1998) report that incubator funds remain small while private, but once opened, quickly increase

in size and revert to median performance5. This strategic behaviour by the investment

management industry provides the appearance of superior performance, with poor-performing

incubator funds (and their track record) closed or merged into a larger fund. Researchers have

also recently found evidence rejecting performance persistence outside traditional equity funds.

Using a sample of hedge funds in the United States, Brown, Goetzmann and Ibbotson (1999)

report no evidence of performance predictability on a raw return and risk-adjusted basis.

Troutman (1991) describes the reliance investors place on past performance data when selecting

funds as a “cognitive error, as many (trustees) see strong past performance and prestigious client

lists as representative of future investment management ability (pp. 37).” The implication of

findings largely supportive of the efficient market hypothesis by this second group of researchers

is neatly summarised by Lakonishok et al (1992). Lakonishok et al (1992) deduce that no

evidence of return persistence over time permits researchers to “make the stronger statement that

not only do (pension) funds on average fail to add value, but the same is true for just about all of

them (pp. 356).”

7

The performance persistence debate has immediate implications for Australia’s superannuation

fund industry, particularly for the choice of fund decision. Shefrin (2000) suggests that “there

does seem to be something of a hot-hands effect. But most investors misread what this

performance says about the future … (investors) tend to attribute too much of that success to

skill rather than luck (pp. 174).” The continued controversy surrounding the predictability (or

otherwise) of fund returns provides the motivation for this study to explore whether the hot-hand

anomaly can be exploited in an economically significant manner for superannuation investors.

In investigating this question, we consider the fund selection problem from the perspective of a

retail investor6. Specifically, this study considers:

1. The randomness of Australian equity superannuation fund performance, using past

performance (raw and risk-adjusted returns) as the criterion for fund selection;

2. A real-world simulation of the actual results achieved by this system of fund selection over

the 1990s; and

3. Implications for the fund selection decision to be made by retail superannuation investors7.

5 Arteaga et al (1998) also find that the first-year success of selection attention funds also attract a large amount of cash inflows, which undermines their subsequent performance. 6 Retail funds are superannuation products that typically have a minimum initial investment amount of AUD 2,000 and subsequent minimum contributions of AUD 100. Retail funds are commonly used by individual-investors with superannuation assets of less than AUD 100,000 to be invested per fund. 7 Alternate questions considered in the contemporary literature have included: comparing professional management versus the returns of individual-investors (Barber and Odean 2000); compensation of advisers (Coles, Suay and Woodbury 2000); investor response to past performance using flow data (Sawicki 2000); and, the price effects of fund trading (Edelen and Warner 2001).

8

3. RESEARCH DESIGN

The data used in this study consists of monthly returns for a sample of 148 retail “Australian

Equity Superannuation Funds - General” as classified by Morningstar, as well as monthly returns

on an accumulation market portfolio index from January 1991 through December 1999. The

fund returns are obtained from Morningstar’s Australian Superannuation Funds database; with

the market return provided by the Australian Stock Exchange. The monthly fund data provided

by Morningstar was net of management fees and excluded entry and exit loads. The sample

included all funds that existed over the sample period (including all terminated funds). The non-

exculsion of funds that did not survive the entire sample period is designed to minimise the

impact of the methodological flaw known as survivorship bias8.

This study is concerned with whether a cognitive bias toward past performance data by

superannuation investors is detrimental to total portfolio returns. Specifically, we ask whether,

on an ex-ante basis, investors are able to differentiate between luck and investment manager skill

in an economically meaningful way. Given our focus on the problem from a retail perspective,

the selection of performance metrics must reflect techniques that are accessible and commonly

employed by individual-investors (and their financial advisers) to guide fund choice. Following

our research motivation, two annual performance metrics are considered: first, raw or risk-

unadjusted returns; and, second, a Sharpe index proxy of risk-adjusted returns, estimated as the

fund’s annual excess return over the Reserve Bank of Australia 13-week treasury note divided by

8 Elton, Gruber and Blake (1996a) argue that samples that do not correct for attrition will overstate the return that funds earn for their investors. Further, ignoring attrition may differentially impact the return reported for funds with different objectives, because funds with different objectives may have different rates of attrition. Brown et al (1992) show that the strength of survivorship bias can be strong enough to account for the evidence favouring return predictability.

9

the fund’s annual standard deviation of returns9. Across the two metrics, several experiments

were conducted to test for persistence in Australian equity superannuation fund returns.

Specifically, the testing procedure was divided into three steps:

1. Following Bogle (1995), we calculated the past returns of all funds, selected the top five,

ten and twenty in each calendar year period, and then recorded the future yearly return

actually achieved10.

2. Grinblatt and Titman (1992), Goetzmann and Ibbotson (1994), Kahn and Rudd (1995)

and Brown et al (1999) test persistence by using a year-by-year cross-sectional regression

of past returns on current returns. Such a technique is also used in this study with a

significant t-statistic for the slope coefficient leading to a rejection of the null hypothesis

that past performance is unrelated to future performance11. A significant positive

(negative) slope coefficient is evidence of performance persistence (reversal); and,

3. Finally, a non-parametric two-way contingency matrix experiment employed by

Goetzmann and Ibbotson (1994), Kahn and Rudd (1995), Malkiel (1995), Bal and Leger

(1996) and Brown et al (1999) is adopted as a confirmatory measure. First, we sort the

funds into winners and losers in period t-1 and winners and losers in period t. We

distinguished winners from losers by ranking fund performance to the median

performance, defining the above-median performers as winners and below-median

performing funds as losers. If the statistical evidence shows that winners in period t-1

9 The Sharpe (1966, 1994) ratio measures the expected return per unit of risk for a zero-investment strategy. Support for selection of the Sharpe index as a proxy for risk-adjusted returns is provided by Bal and Leger (1996) based on Roll’s (1977) critique. Unlike the Sharpe index, Bal and Leger (1996) explain that there is an implicit benchmark portfolio in using the Treynor (1966) and Jensen (1968) techniques. Treynor and Jensen measures can only be estimated with respect to a market index, making it difficult to interpret the measure within a CAPM equilibrium framework due to inefficient benchmarks. For a discussion of the limitations of single-index measures and multi-index alternatives for Australian equity superannuation funds see Drew and Stanford (2000). 10 Due to a sample size of over 800 funds, Bogle (1995) considers the top twenty funds only.

10

persist as winners in period t, we argue that this is evidence of performance persistence12.

The contingency tables illustrate the frequency of four possible outcomes: winner-winner

(WW); loser-winner (LW); winner-loser (WL); and, loser-loser (LL).

4. ANALYSIS

A. Raw returns

We commenced our analysis of performance persistence in raw returns through an examination

of how the best performing funds in one year perform the following year using Bogle’s (1995)

framework. To minimise the possibility of randomness in any single year, we made comparisons

of fund rankings in each year throughout the 1990s (i.e., how the top five, ten and twenty fund

performers of 1991 ranked in 1992, through to how the best performing funds in 1998 performed

in 1999).

11 The cross-sectional regression takes the form αit = α + bαit-1 + ei, where: αit is the performance measure for fund i in period t; b is the slope coefficient measuring performance persistence; αit-1 is the performance measure for fund i in period t-1; and, ei is the random error term.

12 Following Malkiel (1995) the z-test for repeat winners was calculated as follows. Let p be the probability that a winning fund continues to be a winner in the next year, and assume independence across funds. If there is no performance persistence, we would expect p to equal 0.5. Therefore, evidence against persistence in winning would be provided by failing to reject the hypothesis that p = 0.5. Since the random variable Y of the number of persistently winning funds will take the form of a binomial distribution b(n, p), we conduct a binomial test to see if the probability p of persistent winning is greater than 0.5. Malkiel (1995) and Bers (1998) note that when n is reasonably large, say when n ≥ 20, the random Z = (Y – np) / np (1 - p), which is shown in Table 3, will be approximately distributed as normal with mean zero and standard deviation one.

11

Table 1 Rank order of top five, ten and twenty funds, raw returns

Raw returns Rank in 1992 1993 1994 1995 1996 1997 1998 1999 Average Year one Year two 1 68 49 66 132 12 112 127 57 78 2 77 86 33 131 1 109 128 112 85 3 78 9 50 124 7 2 26 113 51 4 62 10 19 94 139 120 27 114 73 5 11 86 18 95 138 1 125 51 66 6 69 87 25 130 135 83 126 55 89 7 70 37 12 128 130 10 20 123 66 8 71 38 13 129 71 5 21 124 56 9 72 85 4 127 34 6 22 120 55 10 66 84 5 125 35 16 24 121 60 11 67 4 53 126 118 15 30 122 67 12 47 35 34 123 119 54 31 126 71 13 48 36 35 121 113 139 25 128 81 14 63 75 37 122 114 138 28 127 88 15 64 76 67 120 115 137 123 130 104 16 74 77 59 118 124 132 124 131 105 17 75 78 60 117 125 44 9 10 65 18 55 79 64 100 121 38 10 13 60 19 56 69 65 99 122 39 11 9 59 20 40 70 37 31 123 40 7 117 58 Summary of average raw returns (% p.a.) Top 5 funds 33.58 -2.22 27.17 -2.92 12.85 17.49 15.31 1.41 25.52 11.83 Top 10 funds 28.67 -3.11 27.72 0.87 10.63 14.20 17.73 5.21 23.03 12.04 Top 20 funds 24.69 -2.32 29.17 -2.88 11.07 11.83 16.53 7.33 23.91 11.83 All funds 14.94 1.11 33.21 -3.97 16.61 13.84 11.76 7.34 31.28 13.90 Market 15.78 -1.40 37.62 -7.78 20.15 12.32 15.17 12.57 33.09 15.22 No. of funds 113 80 87 98 132 143 139 135 135 119

The evidence provided in Table 1 suggests that a top performing fund in one year has borne no

systematic relationship to its ranking in the subsequent year. An equally weighted portfolio of

the top five ranked funds in the first year provides a raw return of +33.58 per cent, over double

the average return for all funds of +14.94 per cent. In the second year, the average return falls to

+11.83 per cent, below the average fund return of +13.90 per cent. Funds that rank in the top

five in a given year, on average, ranked 71 (of 119 funds) in the subsequent year. We follow Lo

12

(1991) and Bogle (1995) in describing this as evidence of mean reversion. When examining the

question of performance persistence over a full decade, it appears from the preliminary analysis

that a strategy of investing in the best performing funds of the past year provides no ex-ante

information regarding the selection of winners in the subsequent year.

Our second test of persistence is a year-by-year cross-sectional regression of past returns on

current returns. Positive estimates of the coefficient b with significant t-statistics are evidence of

persistence. In this case, period t-1 performance contains useful information for predicting

period t performance. Figure 1 shows eight scatter plots (1991-92 through 1998-99) with OLS

lines showing the regression slopes for each of these tests13. These results are summarised in

Table 2.

Table 2 Repeat-winner test results, raw returns

Year b-coefficient t-statistic R2 1991-92 -0.2200 -4.6076 0.2302 1992-93 -0.2465 -3.0253 0.1050 1993-94 -0.0814 -0.9764 0.0111 1994-95 -0.4210 -8.5368 0.4447 1995-96 -0.0455 -0.5192 0.0021 1996-97 0.6326 5.4857 0.1801 1997-98 0.4260 7.1694 0.2787 1998-99 -1.0662 -8.9193 0.3743 1991-92 to 1998-99

-0.1277 -1.7412 0.2033

The results reported in Table 2 show that winners follow winners in an economically meaningful

way in 1996-97 and 1997-98. However, this pattern reverses in 1991-92, 1992-93, 1994-95 and

1998-99. In these periods winners lose, with a significant reversal pattern evident. Over the

13

Figure 1

Year-by-year performance persistence for funds, raw returns

13 Brown et al (1999) note that the upper right quadrant in each panel gives the WW category and the lower left quadrant corresponds to the LL category (referred to in Table 4 of this study).

1991-1992, Raw returns

-1

0

1

2

-4 -2 0 2 4

Period t-1 (1991)

Perio

d t (

1992

)

1995-1996, Raw returns

-1135

-1 1 3 5Period t-1 (1995)

Perio

d t (

1996

)

1992-1993, Raw returns

-1135

-5 0 5Period t-1 (1992)

Perio

d t (

1993

)

1996-1997, Raw returns

-2

0

2

4

-1 1 3 5

Period t-1 (1996)

Perio

d t (

1997

)

1993-1994, Raw returns

-2

0

2

4

-1 4 9

Period t-1 (1993)

Perio

d t (

1994

)

1997-1998, Raw returns

-2-1012

-2 0 2 4

Period t-1 (1997)

Perio

d t (

1998

)

1994-1995, Raw returns

-1

1

3

-2 -1 0 1Period t-1 (1994)

Perio

d t (

1995

)

1998-1999, Raw returns

-1

4

9

-2 -1 0 1 2Period t-1 (1998)

Perio

d t (

1999

)

14

entire period, a non-significant relationship was found, with the estimated slope coefficient being

negative, indicating a bias toward an annual reversal of returns. Given the inconsistency of the

persistence results, we cannot reject the null hypothesis of no systematic persistence on a raw

return basis.

Table 3 Two-way contingency matrix, raw returns

Raw returns Next year % repeat z-test repeat Initial year Winner Loser winners Winners 1991 Winner 12 24 33.3 -2.0 Loser 26 11 1992 Winner 16 26 38.1 -1.5 Loser 26 12 1993 Winner 24 21 53.3 0.4 Loser 20 22 1994 Winner 14 33 29.8 -2.8 Loser 35 11 1995 Winner 38 28 57.6 1.2 Loser 32 33 1996 Winner 39 32 54.9 0.8 Loser 32 36 1997 Winner 44 25 63.8 2.3 Loser 25 41 1998 Winner 25 44 36.2 -2.3 Loser 44 22 1991 to 1999 Winner 213 233 47.8 -0.9 Loser 240 188

Finally, we used contingency tables in a non-parametric test of performance predictability.

Table 3 confirms that there is little evidence of persistence in fund performance over the 1990s.

The null hypothesis of no winning predictability is not rejected in any of the years covered on a

raw returns basis, with no statistically significant results recorded. For the individual years, 5

years out of 9 indicated negative persistence, that is, losing following a winning year. In

addition, a separate analysis was conducted indicating little evidence of persistent under-

15

performance (popularly termed the “cold-hand” phenomenon, or LL) over the sample period14.

The non-parametric evidence suggests that, over the 1990s, winners tended to repeat 48 per cent

of the time, a result largely harmonious with the toss of a fair coin.

B. Risk-adjusted returns

Tucker, Becker, Isimbabi and Ogden (1994) forward that the most egregious error committed

during any assessment of fund performance is conducting a comparison of fund returns without

consideration of differential risk. Further, Tucker et al (1994) observe that while researchers

have been aware of the need to account for differential risk for more than 30 years, individual

investors often persist in ignoring this critical issue. Motivated by the critique of Tucker et al

(1994), we test for predictability in the risk-adjusted returns of Australian equity superannuation

funds using Sharpe indices as a proxy. The reward-to-variability or Sharpe ratio is a popular tool

used by: financial advisers recommending funds to retail investors; asset consultancy firms

providing advice to trustees; and, is the basis of star-rating systems for funds developed by firms

such as Morningstar15.

As with raw returns, Table 4 illustrates that those funds that top league tables on a risk-adjusted

basis in any given year generally fail to outperform industry and market returns in the following

period. The only period when a strategy of investing in the best performing funds (top five, ten

14 We repeat the experiment to test for a cold hand in fund manager returns or persistence in the LL category. A fund that was denoted a loser in the first year, tended to repeat the performance on 44 per cent of occasions over the sample period. Unlike the hot hand results, one significant result was recorded for 1994-95 with a z-test result of -3.5. This indicates significant reversal, that is, winning following losing and vice-versa. This is not evidence of a cold hand, but rather mean reversion. 15 Morningstar rates the investment performance of funds using a rating system of one to five stars. For a complete discussion of the anatomy of the rating system see Blume (1998) and Sharpe (1998). For an analysis of the impact of mutual fund age on Morningstar ratings see Morey (2001).

16

Table 4 Rank order of top five, ten and twenty funds, risk-adjusted returns

Risk-adjusted Rank in 1992 1993 1994 1995 1996 1997 1998 1999 Average Year one Year two 1 72 34 51 133 14 78 135 97 77 2 80 84 66 134 4 139 123 119 94 3 79 30 15 135 9 115 120 120 78 4 66 31 14 112 128 129 125 121 91 5 67 75 10 117 129 119 126 2 81 6 68 76 11 118 130 138 23 89 82 7 69 77 62 119 131 3 130 110 88 8 62 3 39 115 65 2 124 111 65 9 8 68 7 116 109 6 26 102 55 10 59 42 12 122 110 4 27 103 60 11 60 43 13 123 117 5 20 104 61 12 38 44 27 113 118 8 21 105 59 13 39 39 28 114 119 7 22 108 60 14 63 40 29 104 123 44 24 109 67 15 64 41 8 80 124 82 25 113 67 16 45 20 9 101 120 55 30 114 62 17 46 21 5 102 121 93 31 48 58 18 70 17 9 103 122 94 9 49 59 19 71 18 50 91 80 102 10 50 59 20 41 69 67 27 81 103 11 93 62 Summary of average Sharpe ratios (p.a.) Top 5 funds 10.618 -5.826 0.581 3.734 -5.647 2.407 -2.155 -4.168 23.767 1.587 Top 10 funds 8.447 -5.064 1.573 3.858 -5.694 0.243 -0.303 -3.142 19.354 1.353 Top 20 funds 6.419 -4.766 2.202 3.923 -4.577 -0.821 -0.549 -1.659 18.244 1.500 All funds 1.695 -4.097 2.491 2.635 -2.803 0.042 -1.260 -2.147 21.948 2.101 Market 1.977 -3.751 3.317 1.727 -0.738 -0.717 -0.485 -0.349 17.456 2.113 No. of funds 113 80 87 98 132 143 139 135 135 119

and twenty) garnered superior risk-adjusted returns was the 1993 selection-period and 1994

investment-period, with limited evidence of positive persistence also recorded in the second half

of the decade. Across the entire sample period, an investor skilled (or lucky) enough to select

the top five ranked funds each year achieved an average Sharpe index of 10.618 versus the all

fund average of 1.695. Next year, the average Sharpe for the best five performing funds falls to

1.587, which is below the average fund result of 2.101 and the Sharpe ratio for the market

17

portfolio at

18

2.113. The fall from best performing in year t-1 to year t for risk-adjusted returns is dramatic.

Evidence of mean reversion is prevalent with those funds that rank in the top five in a given year

on a risk-adjusted basis, on average, ranked 84 (of 119 funds) in the subsequent year.

Table 5 Repeat-winner test results, risk-adjusted returns

Year Coefficient t-statistic R2 1991-92 0.0020 0.0524 0.0011 1992-93 -0.1066 -1.6023 0.0319 1993-94 -0.1127 1.2486 0.0191 1994-95 -0.4118 -6.4714 0.3152 1995-96 0.0045 0.0678 0.0025 1996-97 0.1294 2.1279 0.0320 1997-98 1.2486 12.0501 0.5219 1998-99 -0.7992 -3.0533 0.0660 1991-92 to 1998-99

-0.0057 0.5525 0.1233

The coefficient, t-statistic and R2 data provided in Table 5 are the result of regressing fund

returns in one year against returns in the next year where returns are reported for funds in both

years16. The risk-adjusted evidence provides no support to the hypothesis of investment

managers having differential skill. In total, statistically significant results were recorded in four

of the eight observation periods, however, an equal proportion of statistically significant positive

and negative results were recorded, suggesting both positive persistence and reversal effects.

Over the sample period, the relationship was not significant, with the estimated coefficient

16 For reasons of space we do not report risk-adjusted scatter diagrams for the eight periods. Summary results from the OLS regressions are provided in Table 5.

19

Table 6 Two-way contingency matrix, risk-adjusted returns

Risk-adjusted Next year % repeat z-test repeat Initial year Winner Loser Winners Winners 1991 Winner 19 29 39.6 -1.4 Loser 16 9 1992 Winner 21 23 47.7 -0.3 Loser 26 10 1993 Winner 19 25 43.2 -0.9 Loser 11 27 1994 Winner 21 23 47.7 -0.3 Loser 38 11 1995 Winner 28 62 31.1 -3.6 Loser 24 18 1996 Winner 28 27 50.9 0.1 Loser 35 49 1997 Winner 49 19 72.1 3.6 Loser 48 19 1998 Winner 41 57 41.8 -1.6 Loser 9 27 1991 to 1999 Winner 226 265 46.0 -1.8 Loser 207 170

suggesting, on average, a slight reversal trend17. Interestingly, across both raw and risk-adjusted

returns, the use of past returns as an explanatory variable in a year-by-year cross-sectional

17 This test does not account for the possibility of cross-correlation among funds. For any given period it is likely that funds managed according to the same “style” will perform similarly, at least to some extent (Malkiel 1995). To test for such potential cross-correlation impacts on our conclusions, we attempted to repeat the experiments for fund categories. However, Morningstar’s classification system does not classify funds within “Australian Equity Superannuation Funds – General” into categories such as growth, income, value, etc. Currently, the only distinguishing features of the funds relate to their names (e.g., ethical, imputation, small companies fund). However, the majority of fund names are simply “Fund Manager X Australian Share Superannuation Fund.” We are currently developing a characteristics based classification system to differentiate manager styles. State-of-the-art research by Davis (2001) directly addresses the issue of whether any particular investment styles reliably deliver abnormal performance and considers whether any evidence of performance persistence can be found when funds of similar styles are compared. For the period 1965 through 1998, Davis (2001) finds that none of the styles employed by US equity mutual fund managers exhibit positive abnormal returns. Davis (2001) reports some evidence of short-run performance persistence among the best-performing growth funds (hot hand) and among the worst performing small-cap funds (cold hand), however, both these results were not sustained beyond one year. Davis (2001) concludes that the impact of cross-correlation among funds is limited, stating that the “economic benefit(s) to active management … are not obvious (pp. 25).”

20

regression fails to capture the cross-section of future returns in an economically meaningful

manner18.

Again we employed a non-parametric test of performance predictability, on this occasion using

risk-adjusted results, as a confirmatory measure. Table 6 shows there are minimal differences

across the sample period in terms of the percentage of persistent winners and what would be

expected by chance. As with the regression results reported in Table 5, significant positive

persistence was recorded around 1997-98. However, a reversal pattern of significance was also

evident from the 1995 (winner) to 1996 (loser). For the individual periods, 7 years out of 9

indicated negative persistence, that is, losing following a winning year. In addition, the data

indicated no evidence of the cold-hand anomaly, with no significant results of loser-loser

repetition in any of the observation periods. Over the 1990s, winners tended to repeat 46 per

cent of the time, a result corroborating the raw return findings.

5. CONCLUDING COMMENTS AND IMPLICATIONS FOR FUND SELECTION

For retail investors faced with the problem of selecting a fund to manage the domestic equity

portion of their asset allocation, there is little likelihood of earning abnormal returns by selecting

the best performing fund managers from the previous period. The evidence presented in this

study supports Bogle’s (1992) claim that “investment management is a field fraught with

fragility and fallibility, where today’s careful, rational fund selections are too often tomorrow’s

embarrassments (pp. 94).” Bogle (1992) notes that while it is virtually impossible to pick the

18 In both the raw and risk-adjusted year-by-year cross-sectional regression experiments, the Durbin-Watson

21

winning funds from year to year, it is easy to pick a single winner – a passive all-market index

fund. The evidence presented in Tables 1 and 4 highlights the superiority of the market portfolio

against a cohort of active funds on a raw and risk-adjusted return basis.

Malkiel (1995) supports this claim, suggesting “most investors would be considerably better off

by purchasing a low expense index fund, than by trying to select an active fund manager who

appears to possess a hot hand (pp. 571).” Recent research by Malkiel and Radisich (2001) finds

that index funds have regularly produced rates of return exceeding those of active equity funds

by 100 to 200 basis points per annum in the United States over the 1990s, finding that there are

two reasons for the excess performance by passive funds: “management fees and trading costs

(pp. 10).”

The issue of fund expenses requires further analysis. The funds investigated in this study had an

average annual management expense ratio of 3.7 per cent per annum19. As discussed in the

research design section of the paper, this study considered fund returns net of management fees

but excluding entry and exit loads to test for return persistence. Therefore, when conducting the

various experiments to test the hot hand anomaly, the costs levied by the investment manager on

entering and exiting their fund was assumed away. The average entry fee for the funds

investigated was 1.8 per cent, with an exit load of 2.0 per cent. These institutional costs are

considerable, and add further weight to the study’s non-rejection of the null hypothesis of no

differential skill among managers.

statistics were 1.98 and 2.01 respectively. Given no evidence of serial correlation, we do not pursue further into the first-order autoregressive AR1 and Augmented Dickey-Fuller (ADF) test. 19 Typically, the management fee is accrued daily and is payable quarterly in arrears (or upon the full withdrawal of the fund) by the redemption of units.

22

The active management techniques employed by the investment managers considered in this

study appear to add little value in the transformation of retirement savings into retirement

income. Active investing is high cost, incurs substantial entry and exit loads and generates

higher taxation burdens for investors than a passive alternative. The marginal benefits (MB) of

active management are far exceeded by its marginal costs (MC). Those funds that can achieve a

resultant MB > MC from active stock selection in any given year seem destined to reverse this

trend the following year. The findings echo Kendall’s (1953) well-known epithet that “the series

looks like a wandering one, almost as if once a week the demon of chance drew a random

number from a symmetrical population of fixed dispersion and added it to the current price to

determine the next period’s price (pp.23).”

Market efficiency survives the challenge from the performance persistence literature. Using a

number of reasonable strategies, the results of this study provide little comfort for those retail

investors (and their financial advisers) relying heavily on a fund’s track record to guide

selection. A rational, self-seeking agent would achieve their retirement income objectives far

more rapidly through implementing a passive approach to both fund and asset selection. As we

proceed toward member-choice in superannuation, the mean-reverting behaviour of investment

manager returns raises a number of questions regarding the optimal design of a superannuation

fund – a topic we consider in our next paper.

23

REFERENCES Arteaga, Kenneth R., Conrad S. Ciccotello, and C. Terry Grant, 1998, “New equity funds:

Marketing and performance.” Financial Analysts Journal, Vol. 54, No. 6, pp. 43-49. Bal, Yasemin, and Lawrence A. Leger, 1996, “The performance of UK investment trusts.”

Service Industries Journal, Vol. 16, No. 1, pp. 67-81. Barber, Brad M., and Terrance Odean, 2000, “Trading is hazardous to your wealth: The common

stock investment performance of individual investors.” Journal of Finance, Vol. 55, No. 2, pp. 773-806.

Bauman, W. Scott, and Robert E. Miller, 1994, “Can managed portfolio performance be predicted?” Journal of Portfolio Management, Vol. 20, No. 4, pp. 31-40.

Bauman, W. Scott, and Robert E. Miller, 1995, “Portfolio performance rankings in stock market cycles.” Financial Analysts Journal, Vol. 51, No. 2, pp. 79-87.

Bers, Martina K., 1998, “Casual relations among stock returns, inflation: Persistence of international mutual fund performance.” Global Finance Journal, Vol. 9, No. 2, pp. 225-240.

Blume, Marshall E., 1998, “An anatomy of Morningstar ratings.” Financial Analysts Journal, Vol. 54, No. 2, pp. 19-27.

Bogle, John C., 1992, “Selecting equity mutual funds.” Journal of Portfolio Management, Vol. 18, No. 2, pp. 94-100.

Brown, Stephen J, William N. Goetzmann, and Roger G. Ibbotson, 1999, “Offshore hedge funds: Survival and performance, 1989-95.” Journal of Business, Vol. 72, No. 1, pp. 91-117.

Brown, Stephen J, William N. Goetzmann, Roger G. Ibbotson, and Stephen A. Ross, 1992, “Survivorship bias in performance studies.” Review of Financial Studies, Vol. 5, No. 4, pp. 553-580.

Brown, Stephen J. and William N. Goetzmann, 1995, “Performance persistence.” Journal of Finance, Vol. 50, No. 2, pp. 679-698.

Carhart, Mark M., 1997, “On persistence in mutual fund performance.” Journal of Finance, Vol. 52, No. 1, pp. 57-82.

Carpenter, Jennifer N., and Anthony W. Lynch, 1999, “Survivorship bias and attrition effects in measures of performance persistence.” Journal of Financial Economics, Vol. 54, No. 3, pp. 337-374.

Cheng, Louis T. W., Lynn K. Pi, and Don Wort, 1999, “Are there hot hands among mutual fund houses in Hong Kong?” Journal of Business Finance and Accounting, Vol. 26, Nos. 1 & 2, pp. 103-135.

Coles, Jeffrey L., Jose Suay, and Denise Woodbury, 2000, “Fund advisor compensation in closed-end funds.” Journal of Finance, Vol. 55, No. 3, pp. 1385-1414.

Davis, James L., 2001, “Mutual fund performance and manager style.” Financial Analysts Journal, Vol. 57, No.1, pp. 19-27.

Drew, Michael E., and James E. Noland, 2000, “EMH is alive and well.” JASSA Journal of the Securities Institute of Australia, No. 4, pp. 15-18.

Drew, Michael E., and Jon D. Stanford, 2000, “Returns from investing in Australian equity superannuation funds, 1991 to 1999.” Discussion Paper, No. 279, (Department of Economics, University of Queensland).

Drew, Michael E., and Jon D. Stanford, 2001a, “Asset selection and superannuation fund performance: A note for trustees.” Economic Papers, Vol. 20, No. 1, pp. 57-65.

24

Drew, Michael E., and Jon D. Stanford, 2001b, “The impact of fund attrition on superannuation returns.” Economic Analysis and Policy, Vol. 31, No. 1, pp. 25-32.

Drew, Michael E., and Jon D. Stanford, 2003, “Returns from investing in Australian equity superannuation funds, 1991 to 1999.” Service Industries Journal, Vol. 23, No. 3, forthcoming.

Edelen, Roger M., and Jerold B. Warner, 2001, “Aggregate price effects of institutional trading: A study of mutual fund flow and market returns.” Journal of Financial Economics, Vol. 59, No. 2, pp. 195-220.

Elton, Edwin J., Martin J. Gruber, and Christopher R. Blake, 1996a, “Survivorship bias and mutual fund performance.” Review of Financial Studies, Vol. 9, No. 4, pp. 1097-1120.

Elton, Edwin J., Martin J. Gruber, and Christopher R. Blake, 1996b, “The persistence of risk-adjusted mutual fund performance.” Journal of Business, Vol. 69, No. 2, pp. 133-157.

Fama, Eugene F., 1970, “Efficient capital markets: A review of theory and empirical work.” Journal of Finance, Vol. 25, No. 2, pp. 383-417.

Goetzmann, William N., and Roger G. Ibbotson, 1994, “Do winners repeat? Patterns in mutual fund behavior.” Journal of Portfolio Management, Vol. 20, No. 2, pp. 9-18.

Grinblatt, Mark, and Sheridan Titman, 1989, “Mutual fund performance: An analysis of quarterly portfolio holdings.” Journal of Business, Vol. 62, No. 2, pp. 394-415.

Grinblatt, Mark, and Sheridan Titman, 1992, “The persistence of mutual fund performance.” Journal of Finance, Vol. 47, No. 5, pp. 1977-1984.

Grossman, Stanford J., and Joseph E. Stiglitz, 1980, “On the impossibility of informationally efficient markets.” American Economic Review, Vol. 70, No. 3, pp. 393-408.

Gruber, Martin J., 1996, “Another puzzle: The growth in actively managed mutual funds.” Journal of Finance, Vol. 51, No. 3, pp. 783-810.

Hendricks, Darryll, Jayendu Patel, and Richard Zeckhauser, 1993, “Hot-hands in mutual funds: Short-run persistence of relative performance, 1974-1988.” Journal of Finance, Vol. 48, No. 1, pp. 93-130.

Jensen, Michael C., 1968, “The performance of mutual funds in the period 1945-1964.” Journal of Finance, Vol. 23, No. 2, pp. 389-416.

Kahn, Ronald N., and Andrew Rudd, 1995, “Does historical performance predict future performance?” Financial Analysts Journal, Vol. 51, No. 6, pp. 43-52.

Kendall, Maurice G., 1953, “The analysis of economic time series - part I: Prices.” Journal of the Royal Statistical Society, Vol. 96, No. 1, pp. 11-25.

Kosowski, Robert, Allan Timmermann, Russ Wermers and Hal White, 2001, “Can mutual fund stars really pick stocks? New evidence from a bootstrap analysis.” Working Paper.

Lakonishok, Josef, Andrei Shleifer and Robert W. Vishny, 1992, “The structure and performance of the money management industry.” Brookings Papers: Microeconomics, No. 1, pp. 339-392.

Leger, Lawrence A., 1997, “UK investment trusts: performance, timing and selectivity.” Applied Economics Letters, Vol. 4, No. 2, pp. 207-210.

Lo, Andrew W., 1991, “Long-term memory in stock market prices.” Econometrica, Vol. 59, No. 5, pp. 1279-1313.

Malkiel, Burton G., 1995, “Returns from investing in equity mutual funds 1971 to 1991.” Journal of Finance, Vol. 50, No. 2, pp. 549-572.

Malkiel, Burton G., and Aleksander Radisich, 2001, “The growth of index funds and the pricing of equity securities.” Journal of Portfolio Management, Vol. 26, No. 2, pp. 9-21.

Morey, Matthew, 2001, “Mutual fund age and Morningstar rating.” Financial Analysts Journal, forthcoming.

25

Moskowitz, Tobias J., 2000, “Discussion.” Journal of Finance, Vol. 55, No. 4, pp. 1655-1695. Roll, Richard, 1977, “A critique of the asset pricing theory’s tests.” Journal of Financial

Economics, Vol. 4, No. 4, pp. 129-176. Sawicki, Julia, 2000, “Investors’ response to the performance of professional fund managers:

Evidence from the Australian wholesale funds market.” Australian Journal of Management, Vol. 25, No. 1, pp. 47-66.

Sawicki, Julia, and Fred Ong, 2000, “Evaluating mutual fund performance using conditional measures: Australian evidence.” Pacific Basin Finance Journal, Vol. 8, Nos. 3 & 4, pp. 505-528.

Sharpe, William F., 1966, “Mutual fund performance.” Journal of Business, Vol. 1, No. 2, pp. 119-138.

Sharpe, William F., 1994, “The Sharpe ratio.” Journal of Portfolio Management, Vol. 21, No. 1, pp. 49-59.

Sharpe, William F., 1998, “Morningstar’s risk-adjusted ratings.” Financial Analysts Journal, Vol. 54, No. 4, pp. 21-33.

Shefrin, Hersh, 2000, Beyond Fear and Greed: Understanding Behavioral Finance and the Psychology of Investing. (Harvard Business School Press: Boston).

Stewart, Scott D., 1998, “Is consistency of performance a good measure of manager skill?” Journal of Portfolio Management, Vol. 24, No. 3, pp. 22-32.

Treynor, Jack L., 1966, “How to rate management of mutual funds.” Harvard Business Review, Vol. 43, No. 1, pp. 63-75.

Troutman, Michael L., 1991, “The Steinbrenner syndrome and the challenge of manager selection.” Financial Analysts Journal, Vol. 47, No. 2, pp. 37-44.

Tucker, Alan L., Kent G. Becker, Michael J. Isimbabi and Joseph Ogden, 1994, Contemporary Portfolio Theory and Risk Management. (West Publishing Company, New York).

Wermers, Russ, 2000, “Mutual fund performance: An empirical decomposition into stock-picking talent, style, transactions costs, and expenses.” Journal of Finance, Vol. 55, No. 4, pp. 1655-1695.